Skip to content

Llama 3.2 Vision 90B

by Meta · llama-3 family

90B

parameters

text-generation vision reasoning multilingual summarization creative-writing

Llama 3.2 Vision 90B is Meta's largest multimodal model, combining powerful text generation with advanced image understanding capabilities. It delivers state-of-the-art performance on visual reasoning, document analysis, chart understanding, and image captioning tasks. With 90 billion parameters and a 128K context window, this model represents the top tier of Meta's vision-language offerings, providing significantly stronger visual comprehension and reasoning compared to the smaller 11B variant.

Quick Start with Ollama

ollama run 90b-q4_K_M
Resources Ollama Hugging Face Official Page
Creator Meta
Parameters 90B
Architecture transformer-decoder
Context 128K tokens
Released Sep 25, 2024
License Llama 3.2 Community License
Ollama llama3.2-vision:90b

Quantization Options

Format File Size VRAM Required Quality Ollama Tag
Q4_K_M rec 45 GB 50 GB 90b-q4_K_M
Q8_0 90 GB 96 GB 90b-q8_0

Compatible Hardware

Q4_K_M requires 50 GB VRAM

Compatible Hardware

HardwareVRAMTypeFitEst. Speed
Mac Studio M4 Ultra 512GB512 GBmacRuns~16 tok/s
Mac Pro M2 Ultra 192GB192 GBmacRuns~16 tok/s
Mac Studio M4 Ultra 192GB192 GBmacRuns~16 tok/s
Mac Studio M4 Max 128GB128 GBmacRuns~11 tok/s
MacBook Pro M4 Max 128GB128 GBmacRuns~11 tok/s
MacBook Pro M5 Max 128GB128 GBmacRuns~11 tok/s
NVIDIA RTX PRO 6000 Blackwell96 GBgpuRuns~38 tok/s
MacBook Pro M3 Max 96GB96 GBmacRuns~8 tok/s
Mac mini M4 Pro 64GB64 GBmacRuns~5 tok/s
Mac Studio M4 Max 64GB64 GBmacRuns~11 tok/s
MacBook Pro M4 Max 64GB64 GBmacRuns~11 tok/s
MacBook Pro M5 Max 64GB64 GBmacRuns~11 tok/s
NVIDIA RTX 6000 Ada Generation48 GBgpuCPU Offload~6 tok/s
NVIDIA RTX A600048 GBgpuCPU Offload~5 tok/s
NVIDIA RTX PRO 5000 Blackwell48 GBgpuCPU Offload~6 tok/s
Mac mini M4 Pro 48GB48 GBmacCPU Offload~2 tok/s
MacBook Pro M3 Max 48GB48 GBmacCPU Offload~2 tok/s
MacBook Pro M4 Pro 48GB48 GBmacCPU Offload~2 tok/s
MacBook Pro M4 Max 48GB48 GBmacCPU Offload~3 tok/s
MacBook Pro M5 Max 48GB48 GBmacCPU Offload~2 tok/s
MacBook Pro M5 Pro 48GB48 GBmacCPU Offload~2 tok/s
Mac Studio M4 Max 36GB36 GBmacCPU Offload~3 tok/s
MacBook Pro M3 Pro 36GB36 GBmacCPU Offload~1 tok/s
MacBook Pro M5 Max 36GB36 GBmacCPU Offload~2 tok/s
83 hardware device(s) cannot run this model at Q4_K_M.

Benchmark Scores

86.0
mmlu